Calling all pro video card hax0rs!! Help me save a GTX 690!

[Polishzx]

I got a hold of an NVIDIA GTX 690 Refrence card. There is a problem with the card where the graphics driver of any machine it is installed on will randomly crash (and recover!) during gameplay.

The card can run furmark for hours, sometimes, without any problem. Other times, it has crashed quickly after starting.

The card will complete maybe 4 out of 5 3dmark benchmark tests, and score very highly in each until randomly the driver crashes in the middle of one.

I feel like there is a small chance the card could be saved with some BIOS modifying trickery?

Does anyone have any ideas on where to begin on how to figure out exactly what is happening as soon as it crashes? There is no graphical corruption when it is working. I am open to any suggestions on how to diagnose and fix this beast!! Thanks for any help! :thumbsup:

 

[thilanliyan]

It probably completes Furmark because nVidia and AMD both throttle their cards while running Furmark.

Have you checked temps? Is your PSU up to scratch?

 

[tviceman]

I'd start with basic stuff like re=seating the graphics card in the PCIe slot. If that doesn't work try it in a different computer if possible. If that doesn't work, try removing the heatsink and re-applying thermal paste. I would not start modifying the bios before trying other things. Worse comes to worse, you can ask the original owner to RMA the card for you as it should still be under warranty.

 

[OCGuy]

1. We need some more specs so that we can make sure it isn't a power issue

2. What are the temps like after a couple passes through 3dmark?

3. What are the temps like when it crashes in game?

4. What driver are you using?

 

[Polishzx]

Gents,

Thanks for the replies. This card has been tested on multiple machines, all of which displayed the same driver crashing issue, so I am very certain it is related to the card itself.

The power supplies on the machines tested were two 750w OCZs, a Dell 1000w and a PC Power & Cooling 610w (capable of 670w max tho). The minimum spec on the GTX 690 is 650w, but it still was able to complete an hour furmark test on the 610w.

As far as temperatures, the hottest it has gotten in Furmark was 84c. I believe the max temperature specified for the 690 is 98c. I already removed the stock sink and put a coating of Arctic Silver 5 on the GPUs.

I don't know what temperatures are during crashes, but they're not too hot as the games are not pushing it whatsoever. The three games I've tested this card with are Unreal Tournament 3, IRacing and Oddworld: Strangers Wrath. Old games, that did not push the card much based on how little the fan throttled up compared to furmark.

I've already tried updating the BIOS to the later versions available on the ASUS and EVGA cards, but the behavior remains the same. I feel like it is something fixable since the card seems to perform just fine the majority of the time and does not crash from being pushed too hard, it just seems random!

Thanks again for any advice.

 

[Polishzx]

Oops, forgot to mention drivers. The most recent ones I'm using are the 327.32 drivers, but I've also tried 320.49 and 320.18.

 

[OCGuy]

Is it actually a NV OEM card? Did you get your hands on a review sample?


Remember, before NV updated the software to make the 690 throttle in Furmark, reviewers were attempting to see if they could make those cards "explode".

You could possibly have a damaged card if that is the case.

 

[Polishzx]

Is it actually a NV OEM cards? Did you get your hands on a review sample?


Remember, before NV updated the software to make the 690 throttle in Furmark, reviewers were attempting to see if they could make those cards "explode".

You could possibly have a damaged card if that is the case.

Yep, it is an NVIDIA OEM card. I don't know what history it has, so it could very well have been previously pushed too hard until damage. However, I feel like if it were truly damaged, it wouldn't occasionally complete long stints of furmark testing. :'(

 

[OCGuy]

Yep, it is an NVIDIA OEM card. I don't know what history it has, so it could very well have been previously pushed too hard until damage. However, I feel like if it were truly damaged, it wouldn't occasionally complete long stints of furmark testing. :'(

Start downclocking to see if it was degraded.

Downclock the cores 100mhz. Increase the Vcore. Downclock the VRAM. (Obviously do all of this separately to isolate)

Although if it were the VRAM you would most likely be seeing it and it would probably hard-lock.

 

[nurturedhate]

Have you tried under clocking the card?